100% on-device
Transcription runs locally on your Mac via Apple's Neural Engine. Your voice never touches a server.
100% local · no account · free forever
Dictate anywhere
on your Mac.
Hold a keyboard shortcut, speak, release. WhisperKey transcribes your voice on-device and pastes the text wherever your cursor is. No cloud, no login, no monthly fee.
Built differently.
No subscription. No accounts. No microphone tap to a server in some other country.
Hold to talk, release to paste
Press your shortcut anywhere — Slack, browser, code editor, terminal. Release and the transcript appears at your cursor.
50+ languages
English by default. Switch to a multilingual model in Settings to dictate in Spanish, French, Hindi, Mandarin, and more.
Free, forever
Open source under MIT. No paywalls, no premium tier, no telemetry. Audit the code, fork it, run it.
Sub-second latency
Built on WhisperKit with Core ML. Short utterances transcribe in under a second on M-series Macs.
Your clipboard, untouched
WhisperKey snapshots your clipboard, pastes the transcript, then restores what you had. You won't lose your copied link.
How it works.
Three steps. Then you forget WhisperKey is there.
-
1
Hold your shortcut
Default is ⌃ ⌥ Space. A subtle waveform appears at the bottom of your screen.
-
2
Speak
The bars react to your voice in real time. Just talk normally — punctuation included.
-
3
Release
The transcript appears at your cursor. No tab-switching, no popup, no copy-paste dance.
Install in 60 seconds.
WhisperKey is a free, open-source app distributed without an Apple Developer ID, so macOS will ask you to confirm the first launch. Here's exactly how.
1
Download & mount
Click the download button — Chrome saves WhisperKey.dmg to your Downloads folder. Double-click it to mount.
2
Drag to Applications
The DMG opens with WhisperKey on the left and an Applications shortcut on the right. Drag the icon across.
3
Open it (the macOS shuffle)
Because the app isn't paid-Apple-Developer-signed, macOS will warn you. Pick whichever path is easier:
4
Grant permissions
WhisperKey's onboarding window walks you through granting Microphone (to record) and Accessibility (to paste at the cursor). It downloads the speech model in the background — about 150 MB — and you're done.
Why the warning?
Apple charges $99/yr for a Developer ID that removes that warning. WhisperKey is free and open-source, so we don't pay it. The code is on GitHub if you want to audit, build, and run it yourself.
What does WhisperKey send anywhere?
Nothing. No analytics, no error reporting, no audio. The only network call ever is the one-time speech model download from HuggingFace on first run. After that it's offline.